Why AI spending growth means more than new investment

AI spending growth requires a closer reading of IT budgets

Gartner’s latest quarterly forecast puts worldwide artificial intelligence (AI) expenditure at $2.7 trillion in 2026, a 49.5% increase, followed by another 36.2% increase in 2027. For a chief information officer (CIO), reading that growth means separating additional investment from changes in existing purchases and rising component costs. All can raise AI expenditure, but each affects the IT budget differently.

Gartner Distinguished VP Analyst John-David Lovelock explains how these changes interact as AI expands alongside other IT activity. He connects three developments: organizations have received additional funding, familiar purchases are gaining AI capabilities, and demand for the components behind those capabilities is raising costs.

Additional funding and changing purchases both expand AI spending

Additional funding offers the most direct route to growth without cuts elsewhere. Lovelock says CIOs received some extra money for AI in 2024 and a smaller addition in 2025, “so there wasn’t a diversion, and now more of their spending is about rebranding than diversion.” In his account, the extra money allowed AI investment to expand alongside existing activities. The later shift concerns how organizations use and classify money already in their budgets.

An ordinary laptop purchase shows how existing expenditure can become AI-related. A company replacing laptops can now buy machines with AI chips, so a routine purchase gains an AI component as the equipment changes. Employees still need computers, and the replacement budget still serves that need. Yet the new specifications bring the purchase into AI expenditure.

Enterprise software can change in the same way as suppliers embed AI in applications companies already buy. The applications gain capabilities while continuing to meet an existing business need. For the budget owner, the next step is to examine what has changed: what the purchase can do, what it costs and what work it supports. Classifying it as AI-related describes a change in the purchase; comparing allocations separately shows whether another activity has lost funding.

Existing work can also become AI-related through a change in purpose. A business-strategy engagement, for example, can shift from examining business strategy to assessing how AI could change it. The organization continues to fund strategy work, but the questions have changed. Lovelock’s description of rebranding therefore includes substantive changes to both purchases and project goals.

These examples show how Lovelock sees new and existing expenditure moving toward AI. “Net new spending is going more towards AI, and existing spending is being transformed towards AI,” he says. He expects that by 2030, “every dollar is going to be an AI dollar in one way or another,” describing AI’s reach across procurement and projects. In his view, AI-related spending grows both through added investment and through activities organizations already fund.

Lovelock draws the same distinction for hyperscalers, the companies operating very large cloud infrastructure. He says they are maintaining conventional cloud expansion while investing in AI: “The hyperscalers haven’t diverted one dime away from their cloud. So AWS, Google, Microsoft, Meta are all continuing to build out their cloud infrastructure at the same rate they were in 2022.” He describes their cloud investment continuing alongside the AI buildout.

Alongside that continuing investment, Lovelock describes the AI buildout as something that “is the largest infrastructure project humanity has ever undertaken,” emphasizing the scale of new physical construction. His account includes substantial infrastructure additions as well as changes to familiar purchases. These place different demands on budgets: construction adds capacity, while AI-enabled replacements and software change what organizations receive through established spending categories.

Category forecasts show where AI spending is growing

These forms of expenditure appear in Gartner’s forecasts for individual AI categories. The growth rates show how quickly spending is expected to expand, but the starting budget determines the dollar increase. A category with faster percentage growth can therefore add fewer dollars than a larger category growing more slowly.

Within this comparison, AI agents and assistants are software systems that perform tasks or help users complete them; Gartner tracks their spending separately from other software. AI development platforms provide tools for building AI applications, and Gartner attributes the upward revision for that category to custom application development by enterprises and software and service providers. As the platforms prove useful, organizations have more reason to buy application-building tools. The revision reflects Gartner’s changing assessment of that demand.

Application development also creates demand for models suited to particular work. Domain-specific language models are designed for a particular field or business use case, and Gartner calls them a “small but growing opportunity” for model providers. That opportunity accompanies its higher growth forecast for generative AI models, which produce content such as text or images. Here too, comparing total expenditure means considering each category’s size alongside its rate of growth.

Memory demand adds pressure to component costs

Spending within these categories also depends on the cost of equipment needed to deliver AI capabilities. Lovelock describes a sequence in which more capable AI servers need more memory and storage, increasing component demand and raising chip prices. Spending can therefore rise through both greater resource requirements and higher prices. A larger bill can reflect the cost of supplying a capability as well as an increase in what the organization can do.

For Lovelock, the unexpected requirement was how much memory the servers needed. He says the increase in spending on their main processing chips had been anticipated. As the servers gained functionality, however, they also needed more memory and storage, which he includes within the wider memory requirement. A procurement plan therefore has to cover the supporting components needed to use the servers’ processing capabilities.

Demand for those components has, in turn, changed the price outlook. Lovelock challenges the usual description of a memory shortage and stresses the strength of demand: “There is a massive over-demand towards memory, and that’s pushed these chip prices incredibly high,” he says. He links server requirements to prices across the component market, which means a project can become more expensive even when its planned scope stays the same.

The price change stands out against Lovelock’s earlier experience of semiconductor forecasting. “Semiconductors for the last 30 years has been a nice, easy market to work through,” he says, describing gradual, comparatively predictable spending increases. In 2015, his forecast put semiconductor expenditure at about a trillion dollars by 2030. He recalls criticism that the projection misunderstood Moore’s law, the long-term pattern of increasing transistor density on chips, and semiconductor fabrication, the process of making chips.

Lovelock now expects chip spending to approach $2 trillion by 2030. He attributes the change to AI, with memory demand and sharply higher chip prices helping to double expected expenditure compared with the outlook before AI. The comparison shows how far his spending expectations have shifted for the underlying technology. It also ties the industry-wide outlook to the components individual buyers need.

For buyers, the practical question is how server requirements and market prices interact over the life of a purchasing plan. A more capable server may need additional memory and storage, while demand from other buyers raises prices for those same components. The budget increase then has two linked causes: the quantity of resources required and the price of each resource. Procurement decisions need to account for both when estimating the cost of a planned capability.

CIOs must reassess capability, cost and risk

This interaction makes it difficult to plan around fixed assumptions about AI’s future capabilities and prices. Lovelock expects several significant changes in a short period: “The next two years, next three years, have at least three major transitions in them.” His warning covers both what organizations will be able to do and what it will cost. A spending decision made today may therefore rest on technical and cost assumptions that change before the investment delivers its expected benefit.

As those assumptions change, CIOs have to revisit the balance between expected benefit and accepted risk. “If you think you know what will happen with AI based on what is happening with AI, you’re wrong,” Lovelock says. He makes the balance between risk and reward the CIO’s “number one priority.” For each new spending commitment, that means assessing the capability it must deliver, the conditions needed to deliver it and the cost the organization is prepared to bear if those conditions change.

Key highlights

• Separate the sources of AI spending growth: CIOs evaluating Gartner’s forecast of 49.5% growth in 2026 need to distinguish new investment, existing purchases gaining AI capabilities and rising prices. Tracking these separately clarifies what additional spending delivers and whether other activities lose funding.
• Compare category growth with budget size: AI agents, models and development platforms have different growth rates and starting budgets. Budget owners need both measures to assess dollar increases and plan allocations.
• Reprice supporting infrastructure: AI servers require more memory and storage, while strong demand raises component prices. Procurement teams need to update both resource quantities and unit costs when estimating the cost of planned capabilities.
• Revisit investment assumptions: Changing capabilities and costs can alter an AI project’s expected return before it delivers benefits. CIOs need to reassess the required capability, delivery conditions and financial exposure as those assumptions change.